Abstract:The diagnosis, prognosis, and treatment of patients with musculoskeletal (MSK) disorders require radiology imaging (using computed tomography, magnetic resonance imaging (MRI), and ultrasound) and their precise analysis by expert radiologists. Radiology scans can also help assessment of metabolic health, aging, and diabetes. This study presents how machine learning, specifically deep learning methods, can be used for rapid and accurate image analysis of MRI scans, an unmet clinical need in MSK radiology. As a … Show more
“…The ECG is a one-dimensional (1-D) signal representing a time series, which can be analyzed using machine learning techniques for automated detection of certain abnormalities. Recently, deep learning techniques have been developed, which provide significant performance in radiological image analysis [4,5]. Convolutional neural networks (CNNs) have recently been shown to work for multi-dimensional (1-D, 2-D, and in certain cases, 3-D) inputs but were initially developed for problems dealing with images represented as two-dimensional inputs [6].…”
The electrocardiogram (ECG) is one of the most extensively employed signals used in the diagnosis and prediction of cardiovascular diseases (CVDs). The ECG signals can capture the heart's rhythmic irregularities, commonly known as arrhythmias. A careful study of ECG signals is crucial for precise diagnoses of patients' acute and chronic heart conditions. In this study, we propose a two-dimensional (2-D) convolutional neural network (CNN) model for the classification of ECG signals into eight classes; namely, normal beat, premature ventricular contraction beat, paced beat, right bundle branch block beat, left bundle branch block beat, atrial premature contraction beat, ventricular flutter wave beat, and ventricular escape beat. The one-dimensional ECG time series signals are transformed into 2-D spectrograms through short-time Fourier transform. The 2-D CNN model consisting of four convolutional layers and four pooling layers is designed for extracting robust features from the input spectrograms. Our proposed methodology is evaluated on a publicly available MIT-BIH arrhythmia dataset. We achieved a state-of-the-art average classification accuracy of 99.11%, which is better than those of recently reported results in classifying similar types of arrhythmias. The performance is significant in other indices as well, including sensitivity and specificity, which indicates the success of the proposed method.
“…The ECG is a one-dimensional (1-D) signal representing a time series, which can be analyzed using machine learning techniques for automated detection of certain abnormalities. Recently, deep learning techniques have been developed, which provide significant performance in radiological image analysis [4,5]. Convolutional neural networks (CNNs) have recently been shown to work for multi-dimensional (1-D, 2-D, and in certain cases, 3-D) inputs but were initially developed for problems dealing with images represented as two-dimensional inputs [6].…”
The electrocardiogram (ECG) is one of the most extensively employed signals used in the diagnosis and prediction of cardiovascular diseases (CVDs). The ECG signals can capture the heart's rhythmic irregularities, commonly known as arrhythmias. A careful study of ECG signals is crucial for precise diagnoses of patients' acute and chronic heart conditions. In this study, we propose a two-dimensional (2-D) convolutional neural network (CNN) model for the classification of ECG signals into eight classes; namely, normal beat, premature ventricular contraction beat, paced beat, right bundle branch block beat, left bundle branch block beat, atrial premature contraction beat, ventricular flutter wave beat, and ventricular escape beat. The one-dimensional ECG time series signals are transformed into 2-D spectrograms through short-time Fourier transform. The 2-D CNN model consisting of four convolutional layers and four pooling layers is designed for extracting robust features from the input spectrograms. Our proposed methodology is evaluated on a publicly available MIT-BIH arrhythmia dataset. We achieved a state-of-the-art average classification accuracy of 99.11%, which is better than those of recently reported results in classifying similar types of arrhythmias. The performance is significant in other indices as well, including sensitivity and specificity, which indicates the success of the proposed method.
“…Moreover, the study did not classify the complete, partial tears of ACL. The study Irmakci et al [ 41 ] was where the average AUC 0.878, 0.857 and 0.859 of models of three classes for AlexNet, ResNet-18 and GoogleNet 0.859 respectively. The one of the state work Tsai et al, 2020 [ 42 ] was used EfficientNet which is optimized and in the case of MRNet the AUC was 0.960, but on the knee, MRI AUC was 0.913 due to imbalanced classes.…”
Section: Discussionmentioning
confidence: 99%
“…The study of [ 40 ] related arthroscopy findings of MRI dataset and used DenseNet architecture upon 489 MRI samples only, in which 163 were from an ACL tear and 245 were from an intact ACL. The comparison study related to musculoskeletal Irmakci et al [ 41 ] performed three CNN architectures AlexNet, ResNet and GoogleNet, that achieved AUC 0.938, 0.956 and 0.890, respectively, detecting ACL tears on MRNet dataset. The ResNet-18 model was found better in the case of an ACL tear, but in the case of abnormalities, the ResNet result was not good.…”
The most commonly injured ligament in the human body is an anterior cruciate ligament (ACL). ACL injury is standard among the football, basketball and soccer players. The study aims to detect anterior cruciate ligament injury in an early stage via efficient and thorough automatic magnetic resonance imaging without involving radiologists, through a deep learning method. The proposed approach in this paper used a customized 14 layers ResNet-14 architecture of convolutional neural network (CNN) with six different directions by using class balancing and data augmentation. The performance was evaluated using accuracy, sensitivity, specificity, precision and F1 score of our customized ResNet-14 deep learning architecture with hybrid class balancing and real-time data augmentation after 5-fold cross-validation, with results of 0.920%, 0.916%, 0.946%, 0.916% and 0.923%, respectively. For our proposed ResNet-14 CNN the average area under curves (AUCs) for healthy tear, partial tear and fully ruptured tear had results of 0.980%, 0.970%, and 0.999%, respectively. The proposing diagnostic results indicated that our model could be used to detect automatically and evaluate ACL injuries in athletes using the proposed deep-learning approach.
“…Two other studies evaluated DL algorithms for meniscus tear detection using the publicly available "MRNet" data set [11]. For meniscus tear detection without differentiating laterality, one study using various CNNs achieved sensitivities of 62-69% and specificities of 76-81% [17]. The other study's CNN achieved a sensitivity of 86% and specificity of 89% by reportedly using coronal T1-weighted MR images only, which, however, are least accurate for radiologists to diagnose meniscal tears [15].…”
Section: Meniscus Tearsmentioning
confidence: 99%
“…Several additional studies using variations of publicly available and custom architecture CNNs reported sensitivities and specificities between 85 and 95% [15][16][17]. Table 1 summarizes key characteristics and performance levels of a group of representative studies.…”
Deep learning-based MRI diagnosis of internal joint derangement is an emerging field of artificial intelligence, which offers many exciting possibilities for musculoskeletal radiology. A variety of investigational deep learning algorithms have been developed to detect anterior cruciate ligament tears, meniscus tears, and rotator cuff disorders. Additional deep learning-based MRI algorithms have been investigated to detect Achilles tendon tears, recurrence prediction of musculoskeletal neoplasms, and complex segmentation of nerves, bones, and muscles. Proof-of-concept studies suggest that deep learning algorithms may achieve similar diagnostic performances when compared to human readers in meta-analyses; however, musculoskeletal radiologists outperformed most deep learning algorithms in studies including a direct comparison. Earlier investigations and developments of deep learning algorithms focused on the binary classification of the presence or absence of an abnormality, whereas more advanced deep learning algorithms start to include features for characterization and severity grading. While many studies have focused on comparing deep learning algorithms against human readers, there is a paucity of data on the performance differences of radiologists interpreting musculoskeletal MRI studies without and with artificial intelligence support. Similarly, studies demonstrating the generalizability and clinical applicability of deep learning algorithms using realistic clinical settings with workflow-integrated deep learning algorithms are sparse. Contingent upon future studies showing the clinical utility of deep learning algorithms, artificial intelligence may eventually translate into clinical practice to assist detection and characterization of various conditions on musculoskeletal MRI exams.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.